Agentic
Like-for-like- GPT-5.4
- 77.2
- Kimi K2.6
- 73.5
- Weighted basis
- 3 vs 3 rows
- Reading
- GPT-5.4 leads
GPT-5.4 vs Kimi K2.6
Updated August 27, 2026. Public scores include evidence status and uncertainty. They are not guarantees for a specific workload.
Decision reading
GPT-5.4 has the higher public score, 73.41 versus 60.14, and the 90% score intervals do not overlap.
19 results are shared. Category rows based on different benchmark sets are marked directional and do not name a winner.
Which one for your work
Recommendations appear only when a shared evidence basis or an explicit operating constraint supports the call. Secondary and unsupported use cases stay disclosed below the initial list.
Agentic work
Tool use, computer use, and multi-step task completion
GPT-5.4
GPT-5.4 leads on the same 3 weighted benchmark rows.
Confidence: stronger
Long documents
Prompts that approach the documented context limit
GPT-5.4
GPT-5.4 has the larger documented context window.
Confidence: documented
Chat turn cost
1K fresh input + 500 output tokens
Kimi K2.6
Kimi K2.6 has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.
Confidence: listed-rates
Show secondary and unsupported calls
Cache-heavy agent loop cost
200K cached + 20K fresh input + 10K output tokens
Kimi K2.6
Kimi K2.6 has the lower estimated token cost for this stated workload. Kimi K2.6 has no published cached-input rate, so cached tokens use its listed input rate.
Confidence: rate-fallback
Repository review cost
50K fresh input + 3K output tokens
Kimi K2.6
Kimi K2.6 has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.
Confidence: listed-rates
Coding work
Code generation, repair, and software-engineering tasks
Not enough matched evidence
The category averages use different weighted benchmark sets, so they are directional rather than like-for-like.
Confidence: limited
What is actually comparable
Shared results can support a head-to-head reading. Results present for only one model describe coverage, not superiority.
18 GPT-5.4 only
19 Shared
13 Kimi K2.6 only
- Shared results
- 19
- GPT-5.4 only
- 18
- Kimi K2.6 only
- 13
- Like-for-like categories
- 2 / 8
3 categories use different evidence sets. Those rows remain visible for coverage context but do not name a winner.
Category results, on a stated basis
Each row states whether both averages use the same weighted benchmark set. Directional and not-comparable rows remain visible, but they never receive a winner in this template.
Knowledge
Like-for-like- GPT-5.4
- 57.6
- Kimi K2.6
- 42.2
- Weighted basis
- 2 vs 2 rows
- Reading
- GPT-5.4 leads
Coding
Directional only- GPT-5.4
- 57.7
- Kimi K2.6
- 64.4
- Weighted basis
- 1 vs 3 rows
- Reading
- Directional only
Math
Directional only- GPT-5.4
- 42.5
- Kimi K2.6
- 67.1
- Weighted basis
- 2 vs 4 rows
- Reading
- Directional only
Multimodal
Directional only- GPT-5.4
- 73.2
- Kimi K2.6
- 79.8
- Weighted basis
- 3 vs 2 rows
- Reading
- Directional only
Reasoning
Not comparable- GPT-5.4
- 74.0
- Kimi K2.6
- Not measured
- Weighted basis
- 1 vs 0 rows
- Reading
- Not comparable
Multilingual
Not comparable- GPT-5.4
- Not measured
- Kimi K2.6
- Not measured
- Weighted basis
- 0 vs 0 rows
- Reading
- Not comparable
Instruction following
Not comparable- GPT-5.4
- Not measured
- Kimi K2.6
- Not measured
- Weighted basis
- 0 vs 0 rows
- Reading
- Not comparable
| Category | GPT-5.4 | Kimi K2.6 | Weighted basis | Reading |
|---|---|---|---|---|
| Agentic | 77.2 | 73.5 | Like-for-like3 vs 3 rows | GPT-5.4 leads |
| Knowledge | 57.6 | 42.2 | Like-for-like2 vs 2 rows | GPT-5.4 leads |
| Coding | 57.7 | 64.4 | Directional only1 vs 3 rows | Directional only |
| Math | 42.5 | 67.1 | Directional only2 vs 4 rows | Directional only |
| Multimodal | 73.2 | 79.8 | Directional only3 vs 2 rows | Directional only |
| Reasoning | 74.0 | Not measured | Not comparable1 vs 0 rows | Not comparable |
| Multilingual | Not measured | Not measured | Not comparable0 vs 0 rows | Not comparable |
| Instruction following | Not measured | Not measured | Not comparable0 vs 0 rows | Not comparable |
Shape of the matched evidence
Only shared public evidence is shown. Sparse evidence stays a ruled list rather than being closed into a radar shape.
Too few matched category axes support a radar. The ruled list below shows only shared benchmark results; positions use each benchmark’s normalized display scale when available.
HLE
Knowledge
FrontierMath v2 (Tier 4)
Math
FrontierMath v2 (Tiers 1-3)
Math
Terminal-Bench 2.0
Agentic
CharXiv
Multimodal
What each workload costs
Three fixed token mixes turn per-token rates into comparable decisions. Each scenario states context fit and whether cached input had to fall back to the published list-input rate.
Chat turn
1K fresh input + 500 output tokens
- GPT-5.4
- $0.01
- Fits in one request
- Kimi K2.6
- $0.00295
- Fits in one request
Kimi K2.6 has the lower modeled cost
Costs use the listed standard API rates.
Repository review
50K fresh input + 3K output tokens
- GPT-5.4
- $0.17
- Fits in one request
- Kimi K2.6
- $0.0595
- Fits in one request
Kimi K2.6 has the lower modeled cost
Costs use the listed standard API rates.
Cache-heavy agent loop
200K cached + 20K fresh input + 10K output tokens
- GPT-5.4
- $0.25
- Fits in one request
- Kimi K2.6
- $0.249
- Fits in one request
- Cached input priced at the published list-input rate
Kimi K2.6 has the lower modeled cost
Kimi K2.6 has no published cached-input rate, so cached tokens use its listed input rate.
Specification differences
Sourced differences are shown directly. Missing facts stay explicit instead of being inferred from a model name or family.
SpecificationGPT-5.4Kimi K2.6
Context window
Maximum documented context; output-token limits may be lower.
GPT-5.4
1.05M
OpenAI pricingKimi K2.6
256K
API model ID
GPT-5.4
gpt-5.4
OpenAI pricingKimi K2.6
Not sourced
Cached-input rate
A missing cached-input rate falls back to the listed input rate only in the stated workload estimate.
GPT-5.4
$0.25 per 1M cached input tokens
OpenAI pricingKimi K2.6
Not published
Documented inputs
GPT-5.4
Not sourced
Kimi K2.6
Not sourced
Documented outputs
GPT-5.4
Not sourced
Kimi K2.6
Not sourced
Provider availability
GPT-5.4
Not sourced
Kimi K2.6
Not sourced
Reasoning profile
GPT-5.4
Reasoning
Kimi K2.6
Reasoning
Weight access
GPT-5.4
Proprietary
Kimi K2.6
Open Weight
License
GPT-5.4
Proprietary
Kimi K2.6
Open Weight
Release date
GPT-5.4
2026-03-05
Kimi K2.6
2026-04-20
If you already use one of these models
- Deployment change
- The models list different providers, so authentication, endpoint behavior, limits, and feature support may change.
- Quality signal
- GPT-5.4 has the higher public score, 73.41 versus 60.14, and the 90% score intervals do not overlap.
- Workload cost
- Repository review: $0.17 vs $0.0595. Cache-heavy agent loop: $0.25 vs $0.249.
- Context tradeoff
- GPT-5.4 has the larger documented window (1.05M).
Run the same representative tasks against both endpoints before changing production traffic.
Self-host vs API cost
Estimates at 50,000 req/day · 1000 tokens/req average.
GPT-5.4
API / mo$13,125
Self-host / moNot listed
Break-even—
Proprietary model — self-hosting not applicable.
Kimi K2.6
API / mo$3,713
Self-host / mo$18,221
Break-even326M/day
Benchmark evidence
The full public result ledger is available for audit without forcing a wide desktop table onto a phone.
Browse raw public benchmark evidence50 rows
Agentic
Terminal-Bench 2.0
GPT-5.475.1%SourceKimi K2.666.7%SourceGPT-5.4 leads this result
CyberGym
GPT-5.479.0%SourceKimi K2.6—Not directly comparable
BrowseComp
GPT-5.482.7%SourceKimi K2.683.2%SourceKimi K2.6 leads this result
OSWorld-Verified
GPT-5.475%SourceKimi K2.673.1%SourceGPT-5.4 leads this result
MCP Atlas
GPT-5.470.6%SourceKimi K2.655.9%SourceGPT-5.4 leads this result
Toolathlon
GPT-5.454.6%SourceKimi K2.650%SourceGPT-5.4 leads this result
τ²-bench results
GPT-5.498.9%SourceKimi K2.6—Not directly comparable
Claw-Eval
Shared sourceGPT-5.460.3%Kimi K2.662.3%Kimi K2.6 leads this result
DeepSearchQA
GPT-5.473.6%SourceKimi K2.692.5%SourceKimi K2.6 leads this result
Gert Labs
Shared sourceGPT-5.464.89%Kimi K2.656.82%GPT-5.4 leads this result
ResearchClawBench
Shared sourceGPT-5.415.3%Kimi K2.618.0%Kimi K2.6 leads this result
JobBench
GPT-5.438.9%SourceKimi K2.6—Not directly comparable
ExploitGym
GPT-5.46.0%SourceKimi K2.6—Not directly comparable
WideResearch
GPT-5.4—Kimi K2.680.8%SourceNot directly comparable
OSWorld 2.0
GPT-5.4—Kimi K2.64.6%SourceNot directly comparable
Coding
LiveCodeBench Pro
GPT-5.487.5%SourceKimi K2.6—Not directly comparable
SWE-bench Pro
GPT-5.457.7%SourceKimi K2.658.6%SourceKimi K2.6 leads this result
React Native Evals
GPT-5.485.3%SourceKimi K2.6—Not directly comparable
Vibe Code Bench
Shared sourceGPT-5.467.42%Kimi K2.637.89%GPT-5.4 leads this result
SWE-bench Verified
GPT-5.4—Kimi K2.680.2%SourceNot directly comparable
LiveCodeBench v6
GPT-5.4—Kimi K2.689.6%SourceNot directly comparable
SWE Multilingual
GPT-5.4—Kimi K2.676.7%SourceNot directly comparable
SciCode
GPT-5.4—Kimi K2.652.2%SourceNot directly comparable
Terminal-Bench 2.0
GPT-5.4—Kimi K2.666.7%SourceNot directly comparable
cursorBench31
GPT-5.4—Kimi K2.647.6%SourceNot directly comparable
Reasoning
Knowledge
GPQA
GPT-5.492.8%SourceKimi K2.690.5%SourceGPT-5.4 leads this result
HLE
GPT-5.452.1%SourceKimi K2.634.7%SourceGPT-5.4 leads this result
HLE w/o tools
GPT-5.439.8%SourceKimi K2.6—Not directly comparable
GPQA-D
GPT-5.492.8%SourceKimi K2.690.5%SourceGPT-5.4 leads this result
HealthBench Hard
GPT-5.440.1%SourceKimi K2.6—Not directly comparable
MedXpertQA (Text)
GPT-5.459.6%SourceKimi K2.6—Not directly comparable
HealthBench Professional
GPT-5.448.1%SourceKimi K2.6—Not directly comparable
Math
FrontierMath v2 (Tiers 1-3)
Shared sourceGPT-5.447.600%Kimi K2.638.966%GPT-5.4 leads this result
FrontierMath v2 (Tier 4)
Shared sourceGPT-5.427.100%Kimi K2.614.580%GPT-5.4 leads this result
AIME26
GPT-5.4—Kimi K2.696.4%SourceNot directly comparable
HMMT Feb 2026
GPT-5.4—Kimi K2.692.7%SourceNot directly comparable
MMAnswerBench
GPT-5.4—Kimi K2.686.0%SourceNot directly comparable
Multimodal
MMMU-Pro
GPT-5.481.2%SourceKimi K2.679.4%SourceGPT-5.4 leads this result
OfficeQA Pro
GPT-5.453.2%SourceKimi K2.6—Not directly comparable
MMMU-Pro w/ Python
GPT-5.482.1%SourceKimi K2.680.1%SourceGPT-5.4 leads this result
CharXiv
GPT-5.482.8%SourceKimi K2.680.4%SourceGPT-5.4 leads this result
ERQA
GPT-5.465.4%SourceKimi K2.6—Not directly comparable
SimpleVQA
GPT-5.461.1%SourceKimi K2.6—Not directly comparable
ScreenSpot Pro
GPT-5.485.4%SourceKimi K2.6—Not directly comparable
ZeroBench
GPT-5.441.0%SourceKimi K2.6—Not directly comparable
MedXpertQA (MM)
GPT-5.477.1%SourceKimi K2.6—Not directly comparable
MathVision
GPT-5.4—Kimi K2.687.4%SourceNot directly comparable
V*
GPT-5.4—Kimi K2.696.9%SourceNot directly comparable
Frequently asked questions
Which is better, GPT-5.4 or Kimi K2.6?
GPT-5.4 has the higher public score, 73.41 versus 60.14, and the 90% score intervals do not overlap. That is the clearest overall quality signal in the current public data.
Which is better for coding, GPT-5.4 or Kimi K2.6?
The current coding averages use different weighted benchmark sets, so BenchLM does not name a winner from them. Read the shared benchmark rows directly and test the models on the same task set.
Which is better for agentic tasks, GPT-5.4 or Kimi K2.6?
GPT-5.4 leads the like-for-like agentic tasks comparison across 3 shared weighted benchmark rows.
Which costs less, GPT-5.4 or Kimi K2.6?
For the stated presets, chat costs $0.01 on GPT-5.4 and $0.00295 on Kimi K2.6; repository review costs $0.17 and $0.0595; the cache-heavy agent loop costs $0.25 and $0.249. Kimi K2.6 has no published cached-input rate, so cached tokens use its listed input rate.
Which has the larger context window, GPT-5.4 or Kimi K2.6?
GPT-5.4 has the larger documented context window: 1.05M, compared with 256K.
Related comparisons
Last updated August 27, 2026
Watch GPT-5.4 vs Kimi K2.6
One weekly email when material rank, price, or benchmark evidence changes make this matchup worth revisiting.
Read a sample issueJoin 2,000+ readers.